trainer_log.jsonl

{"current_steps": 10, "total_steps": 250, "loss": 1.8804, "learning_rate": 7.2000000000000005e-06, "epoch": 0.08, "percentage": 4.0, "elapsed_time": "0:00:54", "remaining_time": "0:21:55", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 20, "total_steps": 250, "loss": 0.469, "learning_rate": 1.4400000000000001e-05, "epoch": 0.16, "percentage": 8.0, "elapsed_time": "0:01:48", "remaining_time": "0:20:42", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 30, "total_steps": 250, "loss": 0.3463, "learning_rate": 1.9991228300988586e-05, "epoch": 0.24, "percentage": 12.0, "elapsed_time": "0:02:41", "remaining_time": "0:19:45", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 40, "total_steps": 250, "loss": 0.1673, "learning_rate": 1.983571470813386e-05, "epoch": 0.32, "percentage": 16.0, "elapsed_time": "0:03:35", "remaining_time": "0:18:51", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 50, "total_steps": 250, "loss": 0.106, "learning_rate": 1.9488760116444966e-05, "epoch": 0.4, "percentage": 20.0, "elapsed_time": "0:04:29", "remaining_time": "0:17:56", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 60, "total_steps": 250, "loss": 0.1623, "learning_rate": 1.895711760239413e-05, "epoch": 0.48, "percentage": 24.0, "elapsed_time": "0:05:23", "remaining_time": "0:17:03", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 70, "total_steps": 250, "loss": 0.1585, "learning_rate": 1.8251134982782952e-05, "epoch": 0.56, "percentage": 28.0, "elapsed_time": "0:06:16", "remaining_time": "0:16:09", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 80, "total_steps": 250, "loss": 0.192, "learning_rate": 1.7384553406258842e-05, "epoch": 0.64, "percentage": 32.0, "elapsed_time": "0:07:10", "remaining_time": "0:15:15", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 90, "total_steps": 250, "loss": 0.1705, "learning_rate": 1.63742398974869e-05, "epoch": 0.72, "percentage": 36.0, "elapsed_time": "0:08:04", "remaining_time": "0:14:21", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 100, "total_steps": 250, "loss": 0.0652, "learning_rate": 1.5239859059700794e-05, "epoch": 0.8, "percentage": 40.0, "elapsed_time": "0:08:58", "remaining_time": "0:13:27", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 110, "total_steps": 250, "loss": 0.1237, "learning_rate": 1.4003490325568953e-05, "epoch": 0.88, "percentage": 44.0, "elapsed_time": "0:09:52", "remaining_time": "0:12:33", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 120, "total_steps": 250, "loss": 0.1669, "learning_rate": 1.2689198206152657e-05, "epoch": 0.96, "percentage": 48.0, "elapsed_time": "0:10:46", "remaining_time": "0:11:39", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 130, "total_steps": 250, "loss": 0.0606, "learning_rate": 1.1322563902571227e-05, "epoch": 1.04, "percentage": 52.0, "elapsed_time": "0:11:39", "remaining_time": "0:10:46", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 140, "total_steps": 250, "loss": 0.1064, "learning_rate": 9.930187397020385e-06, "epoch": 1.12, "percentage": 56.0, "elapsed_time": "0:12:33", "remaining_time": "0:09:52", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 150, "total_steps": 250, "loss": 0.0548, "learning_rate": 8.539169714375885e-06, "epoch": 1.2, "percentage": 60.0, "elapsed_time": "0:13:27", "remaining_time": "0:08:58", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 160, "total_steps": 250, "loss": 0.1326, "learning_rate": 7.176585431571235e-06, "epoch": 1.28, "percentage": 64.0, "elapsed_time": "0:14:21", "remaining_time": "0:08:04", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 170, "total_steps": 250, "loss": 0.0843, "learning_rate": 5.868955701754584e-06, "epoch": 1.3599999999999999, "percentage": 68.0, "elapsed_time": "0:15:15", "remaining_time": "0:07:10", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 180, "total_steps": 250, "loss": 0.0121, "learning_rate": 4.641732050210032e-06, "epoch": 1.44, "percentage": 72.0, "elapsed_time": "0:16:08", "remaining_time": "0:06:16", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 190, "total_steps": 250, "loss": 0.0227, "learning_rate": 3.5188009893686916e-06, "epoch": 1.52, "percentage": 76.0, "elapsed_time": "0:17:02", "remaining_time": "0:05:22", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 200, "total_steps": 250, "loss": 0.0385, "learning_rate": 2.522019095014683e-06, "epoch": 1.6, "percentage": 80.0, "elapsed_time": "0:17:56", "remaining_time": "0:04:29", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 210, "total_steps": 250, "loss": 0.0295, "learning_rate": 1.6707875928990059e-06, "epoch": 1.6800000000000002, "percentage": 84.0, "elapsed_time": "0:18:49", "remaining_time": "0:03:35", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 220, "total_steps": 250, "loss": 0.0206, "learning_rate": 9.816747359488632e-07, "epoch": 1.76, "percentage": 88.0, "elapsed_time": "0:19:43", "remaining_time": "0:02:41", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 230, "total_steps": 250, "loss": 0.0089, "learning_rate": 4.6809332207053083e-07, "epoch": 1.8399999999999999, "percentage": 92.0, "elapsed_time": "0:20:37", "remaining_time": "0:01:47", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 240, "total_steps": 250, "loss": 0.0576, "learning_rate": 1.400396292949513e-07, "epoch": 1.92, "percentage": 96.0, "elapsed_time": "0:21:31", "remaining_time": "0:00:53", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 250, "total_steps": 250, "loss": 0.0308, "learning_rate": 3.898849596456477e-09, "epoch": 2.0, "percentage": 100.0, "elapsed_time": "0:22:24", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}
{"current_steps": 250, "total_steps": 250, "epoch": 2.0, "percentage": 100.0, "elapsed_time": "0:22:24", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}